Publication number | US8102764 B2 |
Publication type | Grant |
Application number | US 11/630,882 |
PCT number | PCT/EP2004/007103 |
Publication date | Jan 24, 2012 |
Filing date | Jun 30, 2004 |
Priority date | Jun 30, 2004 |
Fee status | Paid |
Also published as | EP1766887A1, US20080034086, WO2006002664A1 |
Publication number | 11630882, 630882, PCT/2004/7103, PCT/EP/2004/007103, PCT/EP/2004/07103, PCT/EP/4/007103, PCT/EP/4/07103, PCT/EP2004/007103, PCT/EP2004/07103, PCT/EP2004007103, PCT/EP200407103, PCT/EP4/007103, PCT/EP4/07103, PCT/EP4007103, PCT/EP407103, US 8102764 B2, US 8102764B2, US-B2-8102764, US8102764 B2, US8102764B2 |
Inventors | Paolo Castelli, Maurizio Bartoli |
Original Assignee | Telecom Italia S.P.A. |
Export Citation | BiBTeX, EndNote, RefMan |
Patent Citations (17), Non-Patent Citations (10), Referenced by (2), Classifications (17), Legal Events (2) | |
External Links: USPTO, USPTO Assignment, Espacenet | |
This application is a national phase application based on PCT/EP2004/007103, filed Jun. 30, 2004, the content of which is incorporated herein by reference.
The invention relates to performance evaluation in communication networks.
The invention is advantageously applicable to performance evaluation in multi-service packet switched networks with Quality of Service (QoS) differentiation. A specific feature of these networks lies in that the different services supported by the network are grouped in a limited set of Classes of Service (CoS), depending on their QoS requirements.
Advanced packet scheduling mechanisms—generally in the form of Generalized Processor Sharing (GPS) mechanisms—are used within the network devices to ensure that data flows belonging to different classes of service are handled with different levels of priority and experience different levels of QoS. Nowadays, a typical example of this type of networks is represented by Differentiated Services (DiffServ) IP networks.
The different scheduling mechanisms adopted within these networks, such as Weighted Fair Queuing (WFQ), Deficit Round Robin (DRR) and Modified Deficit Round Robin (MDRR), are examples of GPS-based mechanisms.
However, the field of applicability of the invention is not limited to this type of networks and extends more generally to any kind of packet switched network based on GPS-like scheduling mechanisms.
Analytical methods (as opposed to simulation and methods based on direct measurements) provide a fast, flexible and cost effective approach to network performance evaluation. This approach can be adopted during the network design phase for predictive performance analysis finalized to network dimensioning and “what if” analysis. The same approach can also be applied during actual operation of the network for real time performance estimation based on indirect measurements.
The intrinsic complexity of multi-service networks makes the exact solution of the analysis problem unfeasible; therefore, only approximated approaches are possible.
Two aspects of the problem are particularly significant:
Concerning the former aspect, services with real time requirements are to be distinguished from so-called “elastic” (i.e., best effort) data services.
Real time services (e.g., voice and video transmission services) do indeed require that each packet be received with low delay and delay jitter. As a consequence, performance for these services is properly evaluated at the packet level, in terms of average and maximum packet delay.
Elastic services, on the contrary, are sensitive to the total time required to receive an application level data unit (e.g., a file, a web page) while being essentially insensitive to the delay experienced by each packet. As a consequence, performance for these services is more properly evaluated at the application level in terms of average throughput available for each individual transmission.
Concerning the latter aspect, GPS-like mechanisms serve different traffic aggregates in parallel and control the quota of capacity guaranteed to each flow through pre-assigned weights. By modifying the weights associated to each flow, it is possible to control the capability of each flow to exploit the server capacity and, finally, the service speed perceived by each flow. In practice, a GPS mechanism operates as a priority-based server where the relative priorities of the different traffic aggregates can be varied in a continuous manner. This is in contrast with a classical, absolute priority server approach, where only three states are possible: lower, equal or higher priority.
This flexibility renders GPS-like mechanisms a powerful tool in guaranteeing different levels of service to the different classes. As indicated, GPS-like mechanisms are however extremely difficult to model.
A basic reference textbook such as L. Kleinrock: “Queueing Systems” Volume 1 and Volume 2; J. Wiley & Sons, 1975 provides a quite exhaustive explanation of queuing theory and includes the analysis of a large number of queuing systems; specifically, Volume 2 includes at pages 166-170 the analysis of Processor Sharing (PS) systems. While giving the expression for the average transmission time in a M/G/1-PS system without priorities, GPS systems are not addressed directly.
The article “Performance of Bandwidth Sharing Mechanisms for Service Differentiation in the Internet” by T. Bonald and J. W. Roberts; Proceedings of 13 ITC Specialist Seminar, September 2000, focuses on GPS systems from the viewpoint of “elastic” services only. Therefore, only the average throughput is evaluated. In addition, this article basically approximates a GPS system by means of a system with absolute preemptive priorities and derives upper and lower bounds for the average throughput in a GPS system; these bounds, however tend to be excessively approximated in general load conditions.
The article “Reduced-Load Equivalence and Induced Burstiness in GPS Queues with Long-Tailed Traffic Flows” by S. Borst, et al., Queueing Systems, Vol. 43, N. 4; April 2003, is one of many papers exploiting the so-called Large Deviation Theory to derive the asymptotic behavior of GPS systems under a variety of assumptions about the distribution of the flow length (exponential distribution, heavy tailed distribution etc.). This approach is oriented to the evaluation of the packet loss probability rather than to the characterization of delay and is therefore of little relevance for performance evaluation as intended in the present context.
Finally, U.S. Pat. No. 5,883,819 describes a method for QoS assessment in a multimedia communication network. The method described was developed in an ATM network context with either full sharing of the transmission capacity by the different CoS or a complete segregation of the different CoS. Therefore, the method described in U.S. Pat. No. 5,883,819 is not adapted to be applied successfully in the context of the invention.
In practice, no arrangement is available in the art that permits an integrated evaluation of performance as experienced e.g. by the two types of service (real time and “elastic” services) identified above in the context of GPS-based networks. Approximated solutions focusing either on real time services or on elastic services are available. However, these prior art solutions approximate the behavior of the GPS mechanism with the behavior of an absolute priority mechanism, which causes most of the peculiarities of these mechanisms to be lost.
The need therefore exists of providing an arrangement adapted to overcome the intrinsic drawbacks of the prior art considered in the foregoing by providing a general approach for evaluating the behavior of GPS-like mechanisms.
According to the present invention, that object is achieved by means of a method having the features set forth in the claims that follow. The invention also relates to a corresponding system, a related network as well as a related computer program product, loadable in the memory of at least one computer and including software code portions for performing the steps of the method of the invention when the product is run on a computer. As used herein, reference to such a computer program product is intended to be equivalent to reference to a computer-readable medium containing instructions for controlling a computer system to coordinate the performance of the method of the invention. Reference to “at least one computer” is evidently intended to highlight the possibility for the present invention to be implemented in a distributed/modular fashion.
The arrangement described herein is essentially based on a modification of the arrangements used heretofore for taking into account the effect of the load offered by the different CoS on the performances experienced by a given CoS. More to the point, the modification is based on the introduction of specific weighting factors that account for the relative priority of the different CoS.
Preferably, the arrangement described herein applies such an approach to different procedures adapted for performance evaluation of real time and elastic services, respectively. The procedures in question can be integrated to a common procedure for evaluating the performances perceived by the whole of the different CoS sharing the capacity of a common server. The arrangement described herein can thus be applied in evaluating network performance as experienced by the different CoS (real time CoS as well as elastic CoS) at any network device where packet buffering and scheduling are performed. Exemplary of such devices are the input and/or output interfaces of network switches, routers or multiplexers.
The arrangement described herein is applicable to any number and/or combination of real time and elastic CoS. In order to provide the best results, the arrangement preferably exploits the knowledge of the first and second order statistics of the packet lengths as well as the flow lengths. The arrangement described herein can also be applied, with a greater approximation, when only the first order statistics are known.
The arrangement described herein can be applied directly to the evaluation of performance of a network device subject to a given load. Alternatively, it can be used as the basic building block in an iterative algorithm for link capacity dimensioning. The former application mentioned also includes the utilization of arrangement described herein in a network monitoring system for real time performance evaluation. In addition, the arrangement described herein can be used as a building block for end-to-end performance evaluation within a network.
The invention will now be described, by way of example only, with reference to the enclosed figures of drawing, wherein:
A number of abbreviations used throughout this description are listed below together with their associated meanings:
CoS: Class of Service
DRR: Deficit Round Robin
GPS: Generalized Processor Sharing
MDRR: Modified Deficit Round Robin
PS: Processor Sharing
QoS: Quality of Service
WFQ: Weighted Fair Queuing
The exemplary arrangement described herein is primarily intended for use in evaluating the performance experienced by different classes of services (CoS) sharing the capacity of a packet service system/network. Specifically, in the system considered here a scheduling discipline is implemented based on the Generalized Processor Sharing (GPS) algorithm. Hereinafter, such a system will be identified briefly as a “GPS-like” system.
A GPS-like system may represent any input, output or internal interface of a packet-based communication equipment where packets are buffered before being served and the serving discipline implements a GPS-like algorithm. In particular, the GPS-like system may represents an output interface of a DiffServ-based IP router implementing a scheduling discipline based on algorithms such as Weighted Fair Queuing (WFQ), Deficit Round Robin (DRR) or Modified Deficit Round Robin (MDRR).
The exemplary arrangement described herein distinguishes two main types of CoS: those having real time requirements (Real Time CoS) and those exhibiting no real time CoS (Elastic CoS). The arrangement may handle any number of CoS and/or combination of the two types, without restrictions. All CoS are assumed to be handled by the scheduler according to configuration weights, with the only exception of the first class, which can be handled with absolute priority. This feature facilitates the modeling of special schedulers, such as MDRR schedulers, that provide this capability.
The performance parameters considered may differ depending on the type of CoS and depending on whether the CoS is congested or not.
The performance parameters computed for the non-congested real time CoS are:
The performance parameters computed for the non congested elastic CoS are:
The performance parameter computed for the congested CoS (Real Time CoS as well as Elastic CoS) is the fraction of information lost due to congestion (loss probability).
A high level block diagram of the arrangement described herein is shown in
In the exemplary embodiment shown, the inputs to a link performance evaluation block, designated 100, are:
Commercial devices associate different meanings and formats to the scheduler configuration parameters, but the translation from these specific formats into the fraction of link capacity is usually trivial and does not require to be described in detail here.
The outputs from the link performance evaluation block 100 are:
The link performance evaluation block 100 is comprised of two main sub-blocks or units 110 and 120.
The (sub)block 110 (to which the roles of weighting factors computation and CoS stability evaluation are allotted) performs the computation of the weighting factors used by the performance evaluation algorithm and evaluate whether the different CoS are congested (unstable) or not. In addition, it performs also the computation of intermediate data as required by the block 120.
The block 120 (which is allotted the role of CoS performance evaluation) implements the performance evaluation algorithms.
The performance evaluation arrangement described herein implements different approaches for real time and for elastic CoS. For both types of CoS, operation of the arrangement described herein is obtained by adapting to the more general case of GPS systems the results available for strict priority systems. The adaptation involves a modified approach in taking into account the effect of the load offered by the different CoS on the performance experienced by a given CoS.
Specifically, the modification is based on the introduction of specific weighting factors to account for the relative priority of the different CoS.
The key aspects of this generalization are the two related concepts of “virtual load” offered by a given CoS and of relative priority between two CoS.
More in detail, the load offered by CoS r can be identified as the ratio between the offered traffic B(r) and the link capacity M:
The virtual load of class r, ρ′(r), is defined as the ratio of its offered load to its associated scheduler weight:
The virtual load of a given class provides an indication of the service level received by that class: the lower the virtual load, the better the received service.
Consequently, the relative priority of class s with respect to class r, x(s, r), can be defined as the ratio between the Virtual Loads of the two classes:
If the value of x(s, r) is lower than 1, then class s has a priority higher than class r; on the contrary, if the value is higher than 1, then class s has a priority lower than class r.
Performance Experienced by Real Time CoS
The performance experienced by the real time CoS is evaluated at the packet level, and is expressed in terms of average and maximum packet delay (including the queuing delay and the transmission time). Each packet is transmitted at the full link speed: therefore, at this level of observation, there is no sharing of the server capacity and the system behaves like a classical M/G/1 (Poisson arrival process, general distribution of service times, single server, infinite buffer) queuing system with absolute non preemptive priorities.
The average packet delay experienced by CoS r in an M/G/1 queuing system with absolute non pre-emptive priorities can be computed as:
where:
Basically:
The basic improvement introduced consists in modifying the computation of two terms expressed by formulas (6) and (7) in order to take into account the peculiarities of GPS systems. In practice, two weighting factors α_{1}(s,r) and α_{2}(s,r) are introduced, so that the contribution of class s to the performance perceived by class r can be varied gradually depending on the relative priority of the two classes.
With this modification, the average packet delay experienced by class r can be computed as:
where:
The two summations are computed on the full set of CoS. α_{1}(s,r) and α_{2}(s,r) are the two weighting factors computed on the basis of the virtual loads of the two classes.
The two weighting factors can have any functional form, provided that they satisfy the requirements listed below.
Requirements on α_{1}(s,r), when described as a function of ρ′(s) are:
α_{1}(s,r)=1 for ρ′(s)≦ρ′(r) (i.e., when class s has priority higher than or equal to class r)
α_{1}(s,r) shall be monotonously decreasing to 0 with ρ′(s) when ρ′(s)>ρ′(r) (i.e., when class s has priority lower than class r)
α_{1}(s,r) shall go to zero when ρ′(s) goes to infinite (i.e., when no capacity is allocated to class s), in such a way that the product α_{1}(s,r)·ρ(s) goes to an asymptotic value π_{A}(s,r)=w(s). The exact value of said asymptotic value depends on the global load, but it is limited by the lower bound w(s) and the upper bound w(s)/[w(s)+w(r)].
Requirement on α_{2}(s,r), when described as a function of ρ′(r) are:
α_{2}(s,r)=0 for ρ′(r)≦ρ′(s) (i.e., when class r has priority higher than or equal to class s)
α_{2}(s,r) shall be monotonously increasing with ρ′(r) when ρ′(r)>ρ′(s) (i.e., when class r has priority lower than class s)
α_{2}(s,r) shall asymptotically go to 1 when ρ′(r) goes to infinite (i.e., when no capacity is allocated to class r).
In practice, α_{1}(s,r) and α_{2}(s,r) should preferably have a behavior similar to the one depicted in
A possible functional form for α_{1}(s,r) is, for instance, the exponential expression (11) below, where x(s,r) represents the relative priority of class s with respect to class r as defined by expression (4):
α_{1}(s,r)=Min{1,ρ_{A}(s,r)+(1−ρ_{A}(s,r))·e ^{1-x(s,r)}} (11)
An alternative functional form exhibiting all the desired features is the hyperbolic function (12) below:
In both formulas (11) and (12) the asymptotical value ρ_{A}(s,r) can be approximated as:
On the basis of experimental results, formula (12) seems to provide the best accuracy, as is adopted in the presently preferred embodiment of the arrangement described herein. However, any other functional form satisfying the requirements listed above may be adopted without substantial modification. Similar expressions can be found for the weighting factor α_{2}(s,r).
In principle, the functional form of α_{2}(s,r) can be independent from the form chosen for α_{1}(s,r). However, the accuracy of the model can be improved if some dependency between the two set of weighting factors is introduced. A preferred solution is to determine α_{2}(s,r) according to formula (14) below:
This relationship generally expresses a sort of law of complementarity where the sum of α_{1}(s,r) and α_{2}(s,r) is a value between 1 and 2 which is a function of the load factor Σ_{1}(r).
With this specific function the packet delay averaged over all the different classes appears to be independent of the values of the virtual loads (as it should be) at least for the simple case of two classes.
Formula (8) holds provided that Σ_{1}(r)<1. If this condition is not satisfied, then class r is affected by an overload condition and its packet delay is undefined. In practice, the overload condition will induce the loss of part of the transmitted traffic and the performance degradation can be measured in terms of percentage of lost traffic rather than in terms of packet delay. When this happens, the delay value γ_{PA}(r) remains undefined and the loss probability for the congested class can be estimated with the approach described in the article by S. Borst, et al. cited in the introductory portion of the description. In practice
where:
and S represents the set of non congested classes, i.e., those classes for which Σ_{1}(r)<1.
For non-congested classes, the maximum packet delay (i.e., the ε-quantile of the delay distribution) can be estimated by assuming an exponential distribution. With this assumption we have:
γ_{PM}(r)=γ_{PA}(r)·|ln(s(r))| (17)
Performance Evaluation for Elastic CoS
The performance of the elastic CoS is evaluated at the application level, in terms of average throughput available for the transmission of a data unit or, equivalently, in terms of average transmission time.
When observed at the application level, the queuing system appears to operate essentially like a processor sharing system: application level data units, indeed, are made up of several packets and appear to be served in parallel.
The article by Bonald, et al. cited in the introductory portion of the description provides an expression for the average throughput perceived by a generic flow of class r, derived under the assumption of absolute pre-emptive priorities and hyper-exponential distribution of the data unit size. This formula is reported here for the sake of clarity:
where M is the link capacity, a is the parameter of the hyper-exponential distribution and:
A hyper-exponential distribution with parameter a can be used for approximating a generic distribution with known first and second order moments. It is sufficient to choose for a the value that leads to a match in the second order moment of the generic distribution. In particular, for this application one may choose:
where D_{L }and D_{L2 }are the average document size and the average squared document size, respectively.
This approach is valid provided that all the elastic CoS have the same characteristics in terms of average document size and average squared document size. If this assumption is not verified, then some additional approximations may be introduced by averaging the statistical characteristics of the document size over the different CoS.
The specific case of an exponential distribution is obtained by imposing a=1 and A=1/2.
Formula (18) can be generalized to extend its validity to the case of gradual priorities. The generalization introduced here follows the same approach described for the case of the real time CoS, and consists in substituting the terms Σ_{r }and Σ_{r-1 }with the alternative terms Σ_{1}(r) and Σ_{2}(r) provided by formulas (9) and (10).
With this modification, the new expression for the average throughput becomes:
The average throughput T_{A}(r) represents the average transmission speed experienced during the transmission of a data unit. The average transmission time of a data unit of given size can be determined thereby.
By taking the average document size D_{L}(r), as a reference size, one obtains:
Formula (21) applies provided that Σ_{1}(r)<1. Those CoS for which this condition is not satisfied are congested (unstable) and will experience a non-negligible loss of information. For these CoS the average throughput and the average transmission time are undefined and the performance level is measured in terms of traffic load probability, computed according to expression (15).
More in detail, the block 110 computes in a unit 111 the values of the loads ρ(r) offered by each CoS, and the values of the virtual loads ρ′(r).
The block 110 further processes in a unit 112 the list of the scheduler weights w(r) and generates a new set of weights w′(r) compatible with the presence of a CoS handled with strict priority.
In practice, it generates an artificial weight for the strict priority class and rearranges the remaining weights so that the normalization condition (1) is satisfied. From now on, the modified weights w′(r) will replace on the original weights w(r).
Two further units 113 and 114 compute the values of the parameters Σ_{1}(r) and the values of the parameters Σ_{2}(r). This occurs on the basis of the values w′ (r), ρ(r) and ρ′(r) output from the units 111 and 112.
Additionally, the block 110 further checks (in a unit 115, based on the output of the unit 113) the stability of each CoS and stores the result of the test in the array Flag_{Stab}(r).
Finally, a unit 116 computes the values of parameters a and A.
More in detail, the input data to the block 110 are the following.
The output data are:
As shown, the block 110 preferably includes six main (sub)blocks or units 111 to 116.
The unit 111 performs the computation of the values of the offered loads ρ(r) according to formula (2) and the computation of the values for the virtual loads ρ′(r) according to formula (3).
The unit 112 performs the computation of the modified weights w′(r). This computation depends on the value of Flag_{SP}, according to the following expression:
The unit 113 has the detailed architecture shown in
The unit 114 (having the detailed architecture shown in
The unit 115 assigns the value to Flag_{Stab}(r), depending on the value of Σ_{1}(r):
The detailed architecture of the unit 116 is shown in
The block 120 is devoted to the evaluation of the performances experienced by the different CoS. The block 120 (having the architecture shown in
A first component (module 120.1) evaluates the traffic loss probability P_{Loss}(r) for the congested (unstable) CoS.
A second component (module 120.2) evaluates the average and maximum packet delay for the non congested (stable) Real time CoS.
A third component (module 120.3) evaluates the average throughput and the average transmission time for the non congested (stable) Elastic CoS.
Operation of the modules 120.1 to 120.3 is managed via three modules (that are in fact representative of logical steps) indicated as 120.4 to 120.6.
The module 120.4 scans the CoS list by varying r from 1 to N. The modules 120.5 and 120.6 test the flags Flag_{stab}(r) and Flag_{n}(r) for trueness in order to properly “steer” the procedure over the various alternatives represented by the modules 120.1 to 120.3.
The architecture of the module 120.1 is shown in
The architecture of the module 120.2 is shown in
The inputs to this module are:
The outputs from this module are:
The average packet delay is computed according to the formula (8) in a unit 120.2.1; the maximum packet delay is computed according to the formula (17) in a module 120.2.2.
The architecture of the module 120.3 is shown in
The outputs from the module are:
The average throughput is computed according to the formula (21) in a unit 120.3.1; the average transmission time is computed according to the formula (22) in a unit 120.3.2.
Experimental data demonstrate the improvement achieved by means of the arrangement described herein.
Specifically, in the experiments three CoS have been considered, namely:
The link capacity has been assumed to be M=1.544 kb/s and the global offered load ρ has been varied in the interval [0.7, 0.95]. The adopted scheduler weights have been the following: w(1)=0.7; w(2)=0.2; w(3)=0.1. With these assumptions the Real Time class (CoS 1) has the highest priority and CoS 2 has higher priority than CoS 3.
The observed performance parameters are the average packet delay experienced by the real time class (CoS 1) and the average transmission time experienced by the elastic classes (CoS 2 and CoS 3).
Table 1 below reproduces the basic parameters for the three CoS used in the experiments
Cos | P_{L}(r) | P_{L2}(r) | D_{L}(r) | D_{L2}(r) | Flag_{RT}(r) | ρ(r) |
1 | 0.96 kb | 1 | True | 0.2 ρ | ||
2 | 12.0 kb | 1 | 400 kb | 3 | False | 0.4 ρ |
3 | 12.0 kb | 1 | 400 kb | 3 | False | 0.4 ρ |
Performance of the elastic classes (CoS 2 and CoS 3) has been evaluated by means of a prior art algorithm, based on expression (18), as well as by means of the arrangement described herein, based on expression (21). Similarly, performance of the real time class has been evaluated by means of a prior art algorithm, based on expression (5), as well as by means of the arrangement described, based on expression (8).
The experimental results show an appreciable improvement of accuracy over the prior art for CoS 1 (real time, highest priority) and for CoS 2 (elastic, intermediate priority).
Only a marginal improvement was observed for CoS 3: this is not surprising since, with the set of scheduler weights selected, the level of service provided to the lowest priority class is very close to the level provided in an absolute priority system: for that lowest priority class the advantages of the arrangement described herein are thus less evident
Consequently, without prejudice to the underlying principles of the invention, the details and the embodiments may vary, also appreciably, with reference to what has been described by way of example only, without departing from the scope of the invention as defined in the annexed claims.
Cited Patent | Filing date | Publication date | Applicant | Title |
---|---|---|---|---|
US5381404 * | Jul 12, 1993 | Jan 10, 1995 | Mita Industrial Co., Ltd. | Packet-switching communication network and method of design |
US5883819 | Dec 31, 1996 | Mar 16, 1999 | Northern Telecom Limited | Method and system for quality of service assessment for multimedia traffic under aggregate traffic conditions |
US5917822 * | Apr 12, 1996 | Jun 29, 1999 | Xerox Corporation | Method for providing integrated packet services over a shared-media network |
US6438135 * | Oct 21, 1999 | Aug 20, 2002 | Advanced Micro Devices, Inc. | Dynamic weighted round robin queuing |
US6567415 * | Mar 20, 1999 | May 20, 2003 | Lucent Technologies Inc. | Packet scheduling in a communication network with statistical multiplexing of service classes |
US6657960 * | Nov 23, 1999 | Dec 2, 2003 | International Business Machines Corporation | Method and system for providing differentiated services in computer networks |
US6748433 | Jul 10, 2000 | Jun 8, 2004 | Ectel, Ltd. | Method and system for controlling quality of service over a telecommunication network |
US6768744 * | Mar 21, 2000 | Jul 27, 2004 | Lucent Technologies Inc. | Methods and apparatus for managing communication networks supporting multiple quality of service classes utilizing generalized processor sharing |
US7054267 * | Sep 10, 1999 | May 30, 2006 | Lucent Technologies Inc. | Method and apparatus for scheduling traffic to meet quality of service requirements in a communication network |
US7142513 * | May 23, 2002 | Nov 28, 2006 | Yea-Li Sun | Method and multi-queue packet scheduling system for managing network packet traffic with minimum performance guarantees and maximum service rate control |
US7209437 * | Oct 11, 1999 | Apr 24, 2007 | British Telecommunications Public Limited Company | Computer communication providing quality of service |
US7356631 * | Jan 21, 2005 | Apr 8, 2008 | Himax Technologies, Inc. | Apparatus and method for scheduling requests to source device in a memory access system |
US7447229 * | Oct 27, 2004 | Nov 4, 2008 | Intel Corporation | Method for providing prioritized data movement between endpoints connected by multiple logical channels |
US7768910 * | Feb 3, 2006 | Aug 3, 2010 | Neidhardt Arnold L | Calculations for admission control |
US20020198995 * | Apr 10, 2001 | Dec 26, 2002 | International Business Machines Corporation | Apparatus and methods for maximizing service-level-agreement profits |
US20040017804 * | Jul 19, 2002 | Jan 29, 2004 | Meenaradchagan Vishnu | Arbiter for an input buffered communication switch |
US20050052997 * | Oct 12, 2004 | Mar 10, 2005 | Hector Montes Linares | Packet scheduling of real time packet data |
Citing Patent | Filing date | Publication date | Applicant | Title |
---|---|---|---|---|
US9462025 | Sep 10, 2014 | Oct 4, 2016 | Valens Semiconductor Ltd. | Increasing link throughput to enable admission without exceeding latency variation limits |
US9621612 | Sep 10, 2014 | Apr 11, 2017 | Valens Semiconductor Ltd. | Methods and systems for distributed calculations of latency variation |
U.S. Classification | 370/230.1, 370/229, 370/414, 370/417, 370/230 |
International Classification | H04J1/16, H04L12/54 |
Cooperative Classification | H04L47/2416, H04L12/5693, H04L47/623, H04L47/2408, H04L47/28 |
European Classification | H04L12/56K, H04L47/62D2, H04L47/28, H04L47/24A, H04L47/24B |
Date | Code | Event | Description |
---|---|---|---|
Dec 27, 2006 | AS | Assignment | Owner name: TELECOM ITALIA S.P.A., ITALY Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:CASTELLI, PAOLO;BARTOLI, MAURIZIO;REEL/FRAME:018738/0672 Effective date: 20040823 |
Jul 24, 2015 | FPAY | Fee payment | Year of fee payment: 4 |